GitHub radar
Run a personal AI bot on your own computer, pay only per token
This is an open source app that turns your computer into a personal AI assistant: it works with files, runs commands, and replies in chat or messaging apps, while you connect models through a separate paid gateway called Modelflare.
A personal AI bot that lives on your laptop or server instead of someone else's cloud. It has memory, a work folder, and access to dozens of models through one gateway account.
When it helps
Useful if you want a local assistant for background tasks and chat across apps without locking into one model provider. Not a fit if you work in a terminal all day with one model like Opus — this is built more for chat and files than for development.
Pros
- Open source (MIT), your data and files stay on your own machine
- Access to GPT, Claude, Gemini, Grok, DeepSeek, and more
- Finishes background tasks on its own and notifies you after
- Chat in the app, plus Telegram, Slack, Discord, WeChat, Feishu
Cons
- You pay for every model reply through a Modelflare account
- Not built for heavy terminal-based coding — more chat and files
How to set it up — step by step
- 1Install Docker if you don't have it, then run: docker run -d --name amb -p 3600:3600 -v amb-data:/data ghcr.io/modelflare-ai/anymodel-bot
- 2Get your login token with: docker exec amb cat /data/ui-token
- 3Open http://<your-host>:3600/?token=<token> in a browser and log into the web console
- 4Sign up for a Modelflare account if you don't have one — models won't reply without it
- 5Send the agent the text from the block below to deploy the bot and show you a chat reply
- 6Check that the bot replies in chat, then set its file access permissions
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Deploy AnyModelBot from https://github.com/modelflare-ai/AnyModelBot — via Docker per the README (docker run -d --name amb -p 3600:3600 -v amb-data:/data ghcr.io/modelflare-ai/anymodel-bot), get the token with docker exec amb cat /data/ui-token, open the web console with that token and sign in to Modelflare (ask me if you need a paid account or a model key), then show me how the bot replies in chat.
A regular laptop, no GPU needed — but you'll need a paid Modelflare gateway account for the models, billed per token.
Open on GitHub▌ More finds